All right so are are thirty
finalists today that work with him.
There's also a different conference
this must be the final exam week but
also faculty conference we are so so
she asked me to say the words
introducing them as well and
really I guess what you told
me was a lot like we're told
the first team this was
just a pleasure for
her to work with she said they were
really one of the more professional
the dean of our students and then they
were always coming up with good ideas and
following through and
getting good results out of them of that.
So she was also really thrilled to have
had the opportunity to work with them
that all of them tell you what
they did this illustrates what
I
know the
region is really
just.
Process
was
just trying
to get.
Out
this
time our
trucks.
So we've developed that into
an Excel that all over the world
to determine the best actual strategy for
a given set of input data.
Basically they would input data about
shipments with the same were similar to
our times and from there it's
a two step process in our tool.
First we generate all
the possible that is and
then we use an interview program to
select the best action strategy and
display that to the user so
to clarify the terminology that
just means some number of shipments that
we consider as a single larger shipment
to be picked together whereas a badging
strategy is a combination of that
is such that every shipment that we're
considering has been accounted for and
will be picked exactly once and I also
want to introduce this idea of a maximum
of that size which is simply
limits the number of shipments
that can be included in any given batch
and this is a way for Whirlpool to limit
the number of shipments that were
released to the floor any given time.
So our battery is falling through several
constants that we use approximate the time
it will take to complete that order
including the average time takes to
actually pick the items from the shelves.
The average time it takes to
stage the items at the doctor and
the average time it takes to sort items
back into their respective shipments.
We also have information about the travel
time including the average travel distance
any given skew as well as the average
travel speed of a clamp truck for
the average travel distance and skew
we found that world for any given skew
could have multiple locations in the
warehouse where that you can be found and
we needed an average distance to it in
order to pass me how long it would take
you can see a small example of that here
for this particular studio there are two.
Two locations.
One of which is one hundred
ten feet away here and
the other is fifty feet away here and
let's say your data show that there were
two trips of the fifty foot location and
one trip to the hundred ten foot for
that particular city you would use an
average travel distance of seventy feet.
So once the shipping data has been for
the first step is to generate all
possible that is and that we just mean
any herb site you can see an example
of that if possible that is for
if we have four shipments
here as you can see that
with a maximum that size of one
there are only four possibilities.
Namely each of them by itself.
And that's actually for
choosing one possibilities and for
a maximum that size of two or
the same four possibilities.
Plus force used to for all the different
possible combinations of two shipments and
so if you can see the possible batches for
the other maximum bet sizes.
So once we have all the batches.
We just treat each batch as
a single larger order and
approximate the total amount of
time it will take to complete.
So at this point.
Our goal is to find the combination of
those that it is that minimizes take time
and our initial approach was just to
numerate all the possibilities but
we found out that the problem grew
the scale really quickly the larger
that of the more batteries are for the
more shipments are kind of to consider.
So instead we used in interpreting them to
solve the problem of selecting batching
thread so here you can see the formulation
basically all we're doing here.
The objective function is to minimize the
amount of PIC time needed to pick all of
that is and the only constraint is said
all the shipments have to be accounted for
and picked.
So one foreign step that we took was we
verified and validated our model to make
sure that we were giving something
reliable to the world so verification
means making sure that our code is running
the way we intended it to run and the way
we verified our model was generated option
minutes and solve the problem by hand.
What the best fashion strategy would
be along with what time there would be
according to our model and then we ran
the same options through the tool and
ensure that the results were the same and
for validating what that means is
making sure that our tool is an active
representation of what Orwell actually
does so we can do that we put
individual tasks to your model and
noted how much time the models said it
would take to complete those tasks and
compare that to the historical data
to make sure those were accurate.
So here's an example of our
run run that a run of our all.
So this is the main interface.
Again it's created in Excel.
You just come down and shipments have here
and put your shipment data and this data
is in a format that is easily extracted
from the database that work for users.
So they would just copy paste data
from their database straight into our
Excel St here.
Once you have input the data you come
back and press the Bashevis button and
specify a maximum that size it OK.
And here you can see the results of this
particular write down here you can see
there are three that is for
this run this is the shipment number so
this shipment is by itself and then here's
another that for four shipments and
another with five and you can see the
percent savings time that that represents
and you will not talk about
the deliverables and mentation So
for our project we had to cheat
a little what was the Excel that.
Just ran through the warehouse change
their information up to date
because it's already has this
simple to put the data in and
out of the form they're familiar with
programs which could put
this very difficult problem
actually a very quick run time
constraint with actually a very
large matrix into their programs
over very efficiently as for
the breakdown of the logic or code
that elsewhere so recommendation just
a summary of what went through our project
management project and all sorts of
instructions for using the right tool for
a couple things.
We'll have to do one will likely
have to purchase a warehouse upgrade
which will allow the back and
they will also have to software
upgrade a one time purchase dollars for
every part in their network
a one time purchase as
far as daily operation they have to
run the a couple times throughout
the day whenever they're
orders on the floor and
only impacts the shipping of
this person in each warehouse.
So it does not impact the paper's
interface it's on the truck.
So a very trying one person
product the one below
the World Cup Things happen
that will increase decrease
the average time for
an entire turn decrease
the number of labor
cost drivers as with any process change
for the doctors.
And it's also possible that
the wrong drug doctors will become
so hard to determine
their project in practice
that you know this is not in our
mind all free.
On top of
that size from
three to
now
and there are more time and
there's more right size
five point five percent
and five point five
and.
And I said there
was much higher work and it's
our project rather than that
and our projects and
I've never
met in the first or
well on this project and really
annoyed.
So for the question was about
how long with the information
you give the doctor must
approve the birth so.
Live.
Are you investigated this it
should be about six to eight weeks from
the moment they decide to go forward and
purchase that upgrade and
something else with your version of
that if you put it with them.
What are their insured should just be you
know just start a new software and run.
That's less just a one time install and
which of the other pretty soon within
the cycle and
through an.